Andy in the Cloud

From BBC Basic to Force.com and beyond…


9 Comments

Salesforce DX Integration Strategies

This blog will cover three ways by which you can interact programmatically with Salesforce DX. DX provides a number of time-saving utilities and commands, sometimes though you want to either combine those together or choose to write your own that fit better with your way of working. Fortunately, DX is very open and in fact, goes beyond just interacting with CLI.

If you are familiar with DX you will likely already be writing or have used shell scripts around the CLI, those scripts are code and the CLI commands and their outputs (especially in JSON mode) is the API in this case. The goal of this blog is to highlight this approach further and also other programming options via REST API or Node.js.

Broadly speaking DX is composed of layers, from client side services to those at the backend. Each of these layers is actually supported and available to you as a developer to consume as well. The diagram shown here shows these layers and the following sections highlight some examples and further use cases for each.

DX CLI

Programming via shell scripts is very common and there is a huge wealth of content and help on the internet regardless of your platform. You can perform condition operations, use variables and even perform loops. The one downside is they are platform specific. So if supporting users on multiple platforms is important to you, and you have skills in other more platform neutral languages you may want to consider automating the CLI that way.

Regardless of how you invoke the CLI, parsing human-readable text from CLI commands is not a great experience and leads to fragility (as it can and should be allowed to change between releases). Thus all Salesforce DX commands support the –json parameter. First, let’s consider the default output of the following command.

sfdx force:org:display
=== Org Description
KEY              VALUE
───────────────  ──────────────────────────────────────────────────────────────────────
Access Token     00DR00.....O1012
Alias            demo
Client Id        SalesforceDevelopmentExperience
Created By       admin@sf-fx.org
Created Date     2019-02-09T23:38:10.000+0000
Dev Hub Id       admin@sf-fx.org
Edition          Developer
Expiration Date  2019-02-16
Id               00DR000000093TsMAI
Instance Url     https://customization-java-9422-dev-ed....salesforce.com/
Org Name         afawcett Company
Status           Active
Username         test....a@example.com

Now let’s contrast the output of this command with the –json parameter.

sfdx force:org:display --json
{"status":0,"result":{"username":"test...a@example.com","devHubId":"admin@sf-fx.org","id":"00DR000000093TsMAI","createdBy":"admin@sf-fx.org","createdDate":"2019-02-09T23:38:10.000+0000","expirationDate":"2019-02-16","status":"Active","edition":"Developer","orgName":"afawcett Company","accessToken":"00DR000...yijdqPlO1012","instanceUrl":"https://customization-java-9422-dev-ed.mobile02.blitz.salesforce.com/","clientId":"SalesforceDevelopmentExperience","alias":"demo"}}}

If you are using a programming language with support for interpreting JSON you can now start to parse the response to obtain the information you need. However, if you are using shell scripts you need a little extract assistance. Thankfully there is an awesome open source utility called jq to the rescue. Just simply piping the JSON output through the jq command allows you to get a better look at things…

sfdx force:org:display --json | jq
{
  "status": 0,
  "result": {
    "username": "test-hm83yjxhunoa@example.com",
    "devHubId": "admin@sf-fx.org",
    "id": "00DR000000093TsMAI",
    "createdBy": "admin@sf-fx.org",
    "createdDate": "2019-02-09T23:38:10.000+0000",
    "expirationDate": "2019-02-16",
    "status": "Active",
    "edition": "Developer",
    "orgName": "afawcett Company",
    "accessToken": "00DR000....O1012",
    "instanceUrl": "https://customization-java-9422-dev-ed.....salesforce.com/",
    "clientId": "SalesforceDevelopmentExperience",
    "alias": "demo"
  }
}

You can then get a bit more specific in terms of the information you want.

sfdx force:org:display --json | jq .result.id -r
00DR000000093TsMAI

You can combine this into a shell script to set variables as follows.

ORG_INFO=$(sfdx force:org:display --json)
ORG_ID=$(echo $ORG_INFO | jq .result.id -r)
ORG_DOMAIN=$(echo $ORG_INFO | jq .result.instanceUrl -r)
ORG_SESSION=$(echo $ORG_INFO | jq .result.accessToken -r)

All the DX commands support JSON output, including the query commands…

sfdx force:data:soql:query -q "select Name from Account" --json | jq .result.records[0].Name -r
GenePoint

The Sample Script for Installing Packages with Dependencies has a great example of using JSON output from the query commands to auto-discover package dependencies. This approach can be adapted however to any object, it also shows another useful approach of combining Python within a Shell script.

DX Core Library and DX Plugins

This is a Node.js library contains core DX functionality such as authentication, org management, project management and the ability to invoke REST API’s against scratch orgs vis JSForce. This library is actually used most commonly when you are authoring a DX plugin, however, it can be used standalone. If you have an existing Node.js based tool or existing CLI library you want to embed DX in.

The samples folder here contains some great examples. This example shows how to use the library to access the alias information and provide a means for the user to edit the alias names.

  // Enter a new alias
  const { newAlias } = await inquirer.prompt([
    { name: 'newAlias', message: 'Enter a new alias (empty to remove):' }
  ]);

  if (alias !== 'N/A') {
    // Remove the old one
    aliases.unset(alias);
    console.log(`Unset alias ${chalk.red(alias)}`);
  }

  if (newAlias) {
    aliases.set(newAlias, username);
    console.log(
      `Set alias ${chalk.green(newAlias)} to username ${chalk.green(username)}`
    );
  }

Tooling API Objects

Finally, there is a host of Tooling API objects that support the above features and some added extra features. These are fully documented and accessible via the Salesforce Tooling API for use in your own plugins or applications capable of making REST API calls.  Keep in mind you can do more than just query these objects, some also represent processes, meaning when you insert into them they do stuff! Here is a brief summary of the most interesting objects.

  • PackageUploadRequest, MetadataPackage, MetadataPackageVersion represent objects you can use as a developer to automate the uploading of first generation packages.
  • Package2, Package2Version, Package2VersionCreateRequest and Package2VersionCreateRequestError represent objects you can use as a developer to automate the uploading of second generation packages.
  • PackageInstallRequest SubscriberPackage SubscriberPackageVersion and Package2Member (second generation only) represent objects that allow you to automate the installation of a package and also allow you to discover the contents of packages installed within an org.
  • SandboxProcess and SandboxInfo represent objects that allow you to automate the creation and refresh of Sandboxes, as well as query for existing ones. For more information see the summary at the bottom of this help topic.
  • SourceMember represents changes you make when using the Setup menu within a Scratch org. It is used by the push and pull commands to track changes. The documentation claims you can create and update records in this object, however, I would recommend that you only use it for informationally purposes. For example, you could write your own poller tool to drive code generation based on custom object changes.

IMPORTANT NOTE: Be sure to consider what CLI commands exist to accomplish your need. As you’ve read above its easy to automate those commands and they manage a lot of the complexity in interacting with these objects directly. This is especially true for packaging objects.

Summary

The above options represent a rich set of abilities to integrate and extend DX. Keep in mind the deeper you go the more flexibility you get, but you are also taking on more complexity. So choose wisely and/or use a mix of approaches. Finally worthy of mention is the future of SFDX CLI and Oclif. Salesforce is busy updating the internals of the DX CLI to use this library and once complete will open up new cool possibilities such as CLI hooks which will allow you to extend the existing commands.


24 Comments

Swagger / Open API + Salesforce = LIKE

In my previous blog i covered an exciting new integration tool from Salesforce, which consumes API’s that have a descriptor (or schema) associated with them. External Services allows point and click integration with API’s. The ability for Salesforce to consume API’s complying with API schema standards is a pretty huge step forward. Extending its ability to integrate with ease in a way that is in-keeping with its low barrier to entry development and clicks not code mantra.

swaggerlike

At the time of writing my previous blog, only Interagent schema was supported by External Services. However as of the Winter’18 release this is no longer the case. In this blog i will explore the more widely adopted Swagger / Open API 2.0 standard, using Node.js and Heroku and External Services. As bonus topic, i will also touch on using Swagger Code Generator with Apex!

One of the many benefits of supporting the Swagger / Open API standard is the ability to generate documentation for it. The following screenshot shows the API schema on the left and generated documentation on the right. What is also very cool about this, is the Try this operation button. Give it a try for yourself now!

asciiartswagger

oai

Whats the difference between Swagger and Open API  2.0? This was a question i asked myself and thought i would cover the answer here. Basically as at, Swagger v2.0, there is no difference, the Open API Initiative is a rebranding, born out of the huge adoption Swagger has seen since its creation. This move means its future is more formalised and seems to have more meaningful name. You can read more about this amazing story here.

Choosing your methodology for API development

The schema shown above might look a bit scary and you might well want to just get writing code and think about the schema when your ready to share your API. This is certainly supported and there are some tools that support generation of the schema via JSDoc comments in your code or via your joi schema here (useful for existing API’s).

However to really embrace an API first strategy in your development team i feel you should start with the requirements and thus the schema first. This allows others in your team or the intended recipients to review the API before its been developed and even test it out with stub implementations. In my research i was thus drawn to Swagger Node, a tool set, donated by ApiGee, that embraces API-design-first. Read more pros and cons here. It is also the formal Node.js implementation associated with Swagger.

The following describes the development process of API-design-first.

overview2

(ref: Swagger Node README)

Developing Open API’s with “Swagger Node” 

Swagger Node is very easy to get started with and is well documented here. It supports the full API-design-first development process show in the diagram above. The editor (also shown above) is really useful for getting used to writing schemas and the UI is dynamically refreshed, including errors.

The overall Node.js project is still pretty simple (GitHub repo here), now consisting of three files. The schema is edited in YAML file format (translated to JSON when served up to tools). The schema for the ASCIIArt service now looks like the following and is pretty self describing. For further documentation on Swagger / Open API 2.0 see here.

https://createasciiart.herokuapp.com/schema/
swagger: "2.0"
info:
  version: "1.0.0"
  title: AsciiArt Service
# during dev, should point to your local machine
host: localhost:3000
# basePath prefixes all resource paths 
basePath: /
# 
schemes:
  # tip: remove http to make production-grade
  - http
  - https
# format of bodies a client can send (Content-Type)
consumes:
  - application/json
# format of the responses to the client (Accepts)
produces:
  - application/json
paths:
  /asciiart:
    # binds a127 app logic to a route
    x-swagger-router-controller: asciiart
    post:
      description: Returns ASCIIArt to the caller
      # used as the method name of the controller
      operationId: asciiart
      consumes:
        - application/json
      parameters:
        - in: body
          name: body
          description: Message to convert to ASCIIArt
          schema:
            type: object
            required: 
              - message
            properties:
              message:
                type: string
      responses:
        "200":
          description: Success
          schema:
            # a pointer to a definition
            $ref: "#/definitions/ASCIIArtResponse"
  /schema:
    x-swagger-pipe: swagger_raw
# complex objects have schema definitions
definitions:
  ASCIIArtResponse:
    required:
      - art
    properties:
      art:
        type: string

The entry point of the Node.js app, the server.js file now looks like this…

'use strict';

var SwaggerExpress = require('swagger-express-mw');
var app = require('express')();
module.exports = app; // for testing
var config = {
  appRoot: __dirname // required config
};

SwaggerExpress.create(config, function(err, swaggerExpress) {
  if (err) { throw err; }
  // install middleware for swagger ui
  app.use(swaggerExpress.runner.swaggerTools.swaggerUi());
  // install middleware for swagger routing
  swaggerExpress.register(app);
  var port = process.env.PORT || 3000;
  app.listen(port);
});

Note: I changed the Node.js web server framework from hapi (used in my previous blog) to express. As I could not get the Swagger UI to integrate with hapi.

The code implementing the API has been moved to its asciiart.js file.

var figlet = require('figlet');

function asciiart(request, response) {
    // Call figlet to generate the ASCII Art and return it!
    const msg = request.body.message;
    figlet(msg, function(err, data) {
        response.json({ art: data});
    });
}

module.exports = {
    asciiart: asciiart
};

Note: There is no parameter validation code written here, the Swagger Node module dynamically implements parameter validation for you (based on what you define in the schema) before the request reaches your code! It also validates your responses.

To access the documentation simply use the path /docs. The documentation is generated automatically, no need to manage static HTML files. I have hosted my sample AsciiArt service in Heroku so you can try it by clicking the link below.

https://createasciiart.herokuapp.com/docs/

swaggerui

Consuming Swagger API’s with External Services

The process described in my earlier blog for using the above API via External Services has not changed. External Services automatically recognises Swagger API’s.

externalservicesasciiart

NOTE: There is a small bug that prevents the callout if the basePath is specified as root in the schema. Thus this has been commented out in the deployed version of the schema for now. Salesforce will likely have fixed this by the time you read this.

Swagger Tools

  • SwaggerToolsSwagger Editor, the interactive editor shown in the first screenshot of this blog.
  • Swagger Code Generator, creates server stubs and clients for implementing and calling Swagger enabled API’s.
  • Swagger UI, the browser based UI for generating documentation. You can call this from the command line and upload the static HTML files or use frameworks like the one used in this blog to generated it on the fly.

Can we use Swagger to call or implement API’s authored in Apex?

Swagger Tools are available on a number of platforms, including recently added support for Apex clients. This gives you another option to consume API’s directly in Apex. Its not clear if this is going to a better route than consuming the classes generated by External Services, i suspect it might have some pros and cons tbh. Time will tell!

Meanwhile i did run the Swagger Code Generator for Apex and got this…

public class SwagDefaultApi {
    SwagClient client;

    public SwagDefaultApi(SwagClient client) {
        this.client = client;
    }

    public SwagDefaultApi() {
        this.client = new SwagClient();
    }

    public SwagClient getClient() {
        return this.client;
    }

    /**
     *
     * Returns ASCIIArt to the caller
     * @param body Message to convert to ASCIIArt (optional)
     * @return SwagASCIIArtResponse
     * @throws Swagger.ApiException if fails to make API call
     */
    public SwagASCIIArtResponse asciiart(Map<String, Object> params) {
        List<Swagger.Param> query = new List<Swagger.Param>();
        List<Swagger.Param> form = new List<Swagger.Param>();

        return (SwagASCIIArtResponse) client.invoke(
            'POST', '/asciiart',
            (SwagBody) params.get('body'),
            query, form,
            new Map<String, Object>(),
            new Map<String, Object>(),
            new List<String>{ 'application/json' },
            new List<String>{ 'application/json' },
            new List<String>(),
            SwagASCIIArtResponse.class
        );
    }
}

The code is also generated in a Salesforce DX compliant format, very cool!